181–190 of 240 posts
The problem was the smart phone. With that closed platforms became the standard. Non productive hardware is now ubiquitous, most of it can't be fixed. The open risc-v tablet is the only exit from this mess. Forget ARM and X86 empires, they will only exploit you as a consume only slave. It needs working GPU and swap-able battery though!
Earlier quoted context omitted.
What the fuck are we doing as a society that we have such a system of perverse incentives in place?
Our society is geared -- at all levels -- towards minimizing expense regardless of the impact on quality or even on society in general. We're all racing to the bottom.
Suppose Alice is making $40,000/year and comes up with a way to automate a third of her job. So you start paying her $50,000/year and give her some other work to do. Then Bob and Carol each find a way to automate a third of their own jobs, so now they all make $50,000 and have made Don redundant.
The company is now paying $150,000 in total salary instead of $160,000 and only has to pay insurance and provide office space for three employees instead of four. Meanwhile the workers who found ways to improve efficiency are making more money and have the incentive to do it again and get another raise.
Companies may not actually do this, but those companies are mismanaged and putting themselves at a competitive disadvantage.
The problem was the smart phone. With that closed platforms became the standard. Non productive hardware is now ubiquitous, most of it can't be fixed. The open risc-v tablet is the only exit from this mess. Forget ARM and X86 empires, they will only exploit you as a consume only slave. It needs working GPU and swap-able battery though!
What difference does the ISA make in this respect?
I'm skeptical that is how it would play out in practice though.
Earlier quoted context omitted.
What the fuck are we doing as a society, that this lady, with a machine that could compute the above in a fraction of a second, has to spend a decent chunk of her work time doing this? To put it bluntly: do you want her to be unemployed instead?
Yes. A human mind is a terrible thing to waste. Plenty of people don't look for more fulfilling work because what they already have is "good enough" and it brings home a paycheck that supports the people who depend on them. Oftentimes, people need a push, and doing so can be an act of compassion. Laying people off is not some death sentence; it's a disruptive event that well-adjusted people will recover from. That's…
Some of us are not well adjusted. Some of us are barely hanging on, barely able to manage the demands of our lives and our society.
Being laid off at this point for me would be worse than disruptive it would be an outright disaster.
I question the article's framing of CPUs as "universal" and GPUs as "specialized". In theory, they can both do any computation, so they differ only in their performance characteristics, and the deep learning revolution has shown that there is wide range of practical workloads that is non-viable on CPUs. The reason OpenAI runs GPT-4 on GPUs isn't that it's faster than running it on CPUs--they do it because they _can't…
Earlier quoted context omitted.
Many companies don’t even give top performers raises large enough to cover inflation. Who are you kidding here
I get bonuses because my manager knows very well he has much bigger issues when I leave. You have to have a certain amount of flexibility of course and be willing to quit. I'm still very sure that good performance is overall much more beneficial than staying and acting at the bottom. You are hurting yourself while you do mondaine tasks while the other person gets compound interest and new skills.
Earlier quoted context omitted.
Good question! Wait until you are 50, maybe you can answer it then ;-)
I (and most of my social circle) am over 50, and I can answer that question: there's no solid reason you can't train for and be successful in a different career later in life. I've seen it happen too often to think otherwise. Whether or not you want to is an entirely different question, of course.
Earlier quoted context omitted.
She doesn't have to tell anyone
Exactly my point. But if you are a manager, you might wanna consider if that is the incentive structure you wanna have at your organization: People who do good work get "punished" with more work, people who keep it low don't. I certainly have left jobs because of that.
Earlier quoted context omitted.
I (and most of my social circle) am over 50, and I can answer that question: there's no solid reason you can't train for and be successful in a different career later in life. I've seen it happen too often to think otherwise. Whether or not you want to is an entirely different question, of course.
Not there yet, but slowly getting there. Of course you can train for a different career after you are 50, but you will also have a good idea what kind of career is not a good fit for you (anymore). So just because certain careers are now looking for people, doesn't mean that these are a good fit for you.
Our youth is when the majority of what we do is try things out to see what fits and what doesn't. In our older years, most of that experimentation is behind us and we have a pretty solid idea of what fits us and what doesn't.
The trick is that the amount of experimentation should never be reduced to zero.
I question the article's framing of CPUs as "universal" and GPUs as "specialized". In theory, they can both do any computation, so they differ only in their performance characteristics, and the deep learning revolution has shown that there is wide range of practical workloads that is non-viable on CPUs. The reason OpenAI runs GPT-4 on GPUs isn't that it's faster than running it on CPUs--they do it because they _can't…
> 1. substantial numbers of calculations can be parallelized
> 2. the computations to be done are stable and arrive at regular intervals ('regularity')
> 3. relatively few memory accesses are needed for a given amount of computation ('locality')
Where (1) fails, any modern multicore + SIMD + ILP desktop/console/mobile CPU will run at a tiny fraction of its peak throughput. While sufficiently small serial tasks still complete in "good enough" time, the same could be said of running serial programs on GPU (in fact this is sometimes required in GPU programming). People routinely (and happily) use PL implementations which are ~100x slower than C. The acceptibility of ludicrous under-utilization factors depends on the tininess of your workload and amount of time to kill. Parallelism is used broadly for performance; it's about as un-specialized as you can get!
(2) and (3) are really extensions of (1), but both remain major issues for serial implementations too. There mostly aren't serial or parallel applications, rather it's a factor in algorithm selection and optimization. Almost anything can be made parallel. Naturally you specialize HW to extract high performance, which requires parallelism, for specialized HW as for everywhere else.
The authors somewhat gesture towards the faults of their definition of "specialized" later on. Truly specialized HW trades much (or all) programmability in favor of performance, a metric which excludes GPUs from the last ~15 years:
> [The] specialization with GPUs [still] benefited a broad range of applications... We also expect significant usage from those who were not the original designer of the specialized processor, but who re-design their algorithm to take advantage of new hardware, as deep learning users did with GPUs.